Battery energy storage system (BESS) has great potential to combat global warming. However, internal abnormalities in the BESS may develop into thermal runaway, causing serious safety incidents. In this study, the multiscale information fusion is proposed for thermal abnormality detection and localization in BESSs. We introduce the concept of dissimilarity entropy as a means to identify anomalies for lumped variables, whereas spatial and temporal entropy measures are presented for the detection of anomalies for distributed variables. Through appropriate parameter optimization, these three entropy functions are integrated into the comprehensive multiscale detection index, which outperforms traditional single-scale detection methods. The proposed multiscale statistic has good interpretability in terms of system energy concentration. Battery system internal short circuit (ISC) experiments have demonstrated that our proposed method can swiftly identify ISC abnormalities and accurately pinpoint the problematic battery cells.